stringnetpy - A stringnet python implementation

REFERENCE:
---
TITLE: StringNet as a Computational Resource for Discovering and Investigating Linguistic Constructions
AUTHORS: David Wible - Nai-Lung Tsao
URL: http://j.mp/stringnet


DEFINITION OF HYBRID-GRAM:
---
In StringNet a hybrid-gram is defined by two criteria:
1. it must include at least one lexical gram (which is either a lexeme or a specific word form)
2. it must appear at least 5 times in BNC


DISCUSSION AND PERSONAL CHOICES:
---
1. in Alice corpus there are not lexemes, but lemmas. I decided to consider lemmas instead of lexemes.
According to Wikipedia, lemmas are a particular type of lexemes, so the decision makes sense 
http://en.wikipedia.org/wiki/Lexeme

2. I decided to filter out all sentences that contain one or more punctuation signs

3. To operate vertical pruning I refer to this definition:
"In a word, vertical pruning eliminates hybrid n-grams containing 
POS grams which do not represent attested substitutability.", taken from the paper
I use a threshold of 80% (same of paper)

4. They filter out all hybrid-grams that do not occur at least 5 times in BNC (criterion 2)
I decided to filter out hybrid-grams that do not occur at least 5 times in Alice